Dylan Ponsford

april lab, University of Edinburgh

prof_pic.jpg

I am a second-year PhD student at april lab supervised by Antonio Vergari (primary) and Henry Gouk (secondary). (april lab \(\subset\) Institute for Machine Learning \(\subset\) School of Informatics \(\subset\) University of Edinburgh).

My research focuses on efficient and reliable probabilistic methods, with a particular focus on a class of (tractable!) models called probabilistic circuits. Recently, I have also developed an interest in applications to tabular data. You can explore the intersection of these interests in my recent work A Sobering Look at Tabular Data Generation via Probabilistic Circuits.

Previously, I did my MSc in AI also at Edinburgh, concluding with my master’s dissertation “Quantizing Tractable Probabilistic Models” (also supervised by Antonio!). Before that (an increasingly long time ago), I did my BSc in mathematics at the University of Warwick, with a year abroad at Technische Universität Berlin in Germany.

news

Aug 21, 2026 I attended UAI 2026 and presented a poster of our work TabPC!
Aug 10, 2026 The arXiv version of TabPC has been updated to include comparisons with the tabular foundation model TabPFN-3!
Jun 01, 2026 Our work TabPC (a.k.a. A Sobering Look at Tabular Data Generation via Probabilistic Circuits) has been accepted as a poster at UAI 2026!
Mar 24, 2026 Our pre-print A Sobering Look at Tabular Data Generation via Probabilistic Circuits has been released on arXiv!
Mar 06, 2026 I attended the Dagstuhl seminar Tensor Factorizations Meet Probabilistic Circuits and gave a talk!